Interaction reversal between knowledge-to-be-memorized and ideas-to-be-developed

From Issawiki
Jump to: navigation, search

There's something really deep/strange going on between Anki cards ("knowledge-to-be-memorized") vs spaced writing inbox ("ideas-to-be-developed"), that I haven't seen anyone (not even Andy) talk about. When you review cards in Anki, pressing "good" (meaning you got the answer right) makes it show up later in time, whereas pressing "again" (meaning you got it wrong) makes it show up sooner in time. But with a writing inbox, things are arguably reversed: if you interact with a note a lot, that means you remember it more, but it also means it's an interesting note so you want to keep seeing it to build on it more and more; whereas a bad note is a bad idea so you don't want to be reminded of it very much. In other words, saying "good" (= good idea) should make it come up sooner, and saying "bad" (= bad idea, which sounds like "again") should make it come up later.

"notes identity is very crude right now: we just check the sha1 hash, so any modification to a note will turn it into a note with different identity, which means the review schedule will reset. I think there's a decent chance this is actually ok: the notes you modify are the notes you are actually engaging with, so you actually want them around more frequently."[1]

This suggests that the buttons "Good", "Again", "Hard", "Easy" are not the right abstractions to be using for responses to an SRS prompt. An SRS system that integrated both knowledge-based flashcards (to be memorized) and a writing inbox (to develop an idea over time) would need responses which "feel right" in both contexts. If an SRS is extended even further to other contexts, then we probably also want responses that work in all contexts.

Readwise uses "Never | Soon | Later | Eventually" for its response buttons.[2]

Related idea: so interestingly, if you want to implement like a SRS for note taking, then there are two distinct reasons why you might want to "push back" a note further into the future: (1) you're like "meh, this ideas sucks, don't need to see it for a long time"; (2) you're like "this is a great idea! but I know it very well/remembered it all this time, so I don't need to be reminded about it for a long time".

One idea is to use dual ratings for spaced inbox.

also related is https://github.com/andymatuschak/orbit/issues/168

See also

References

What links here